EP1443687A1 - Improved VCSEL analog optical link - Google Patents

Abstract

An improved analog VCSEL optical link includes thefollowing elements listed in their downstream order from theinput to the output: an RF preamplifier (A1) for receiving abandpass RF signal; a first impedance matching network(LASINT) to match the input impedance of a downstream VCSELlaser to the output impedance of the RF preamplifier (A1);the VCSEL laser coupled to one end of a multimode opticalfibre of given length for transducing optically theelectrical bandpass RF signal; a PIN diode coupled to theother end of the multimode optical fibre for transducing intoelectrical the optical signal coming from the fibre; a secondimpedance matching network (C2, L2) to match the outputimpedance of the PIN diode to the input impedance of adownstream low-noise transimpedance amplifier (T1) andminimise the noise figure of said transimpedance amplifier(T1) in the meanwhile; the transimpedance amplifier (T1)including a low-noise FET in common source configuration; anRF power amplifier (A2); an RF signal envelope detectorgenerating a control signal for a variable phase shifterplaced downstream the RF power amplifier (A2) to compensatefor the AM/PM distortion introduced by both the optical fibreand the power amplifier itself; and, if necessary, a gainlinearisation network of the RF power amplifier (A2) and/orthe RF preamplifier (A1). The analog VCSEL optical link issuitable for the following uses: connecting indoorapparatuses with tower antennas either static, i.e.point-to-point radio relay, or directive arrays, i.e.point-to-multipoint systems; building-up a fibreinfrastructure for connecting centralised base stations withrespective pool of remote antennas serving as many cells of acellular telephone system such as: PCS, GSM, UMTS, andsimilar; intra-building connections for distributing, CATV services (fig.4).

Description

FIELD OF THE INVENTION
• The present invention relates to the field of multimodeanalog optical fibre communications and more precisely to animproved VCSEL analog optical link.
BACKGROUND ART
• Historically, Light Emitting Diodes (LED) haverepresented the most popular light emitting devices in datatransmission through optical fibres, nowadays solid statelasers are replacing LEDs in almost all optical fibrecommunications. Different materials can be mixed to thesilica during the manufacturing process of the optical fibresin order to obtain particular wavelength windows of minimumpropagation losses for the light. Well exploited windows are:the first (around 850 nm), the second (around 1,300 nm), andthe third (around 1,550 nm). All these optical windows belongto the infrared spectral range. Lasers are designed foremitting a coherent light inside said optical windows.Optical fibres can be single-mode or multimode, thesingle-mode have lower core/cladding ratios, typically9-10/125 µm against 50-65/125 µm of the multimode.Single-mode fibres are more expensive than multimode and havethe advantage to exhibit small attenuation of the light andminor intermodal dispersion, so that they are preferred forthe longer distances. Solid state lasers have intrinsicnon-linearities because of the coupling between the carrierdensity and the photon density through the material. Thesenon-linearities generate spurious Intermodulation Distortion(IMD) that can fall within the detection bandwidth, such asthe third order (IMD3) ones. A measure that quantifies theeffect of the non-linearity of the laser is the Spurious Free Dynamic Range (SFDR), defined as the output power range overwhich the modulation signal can be detected above the noisefloor while spurious IMD3 remains below the noise floor. TheSFDR value is measured with standard two tone method andconventionally normalised into 1 Hz bandwidth. The SFDR unitare dB.Hz^2/3 because multiplying X times the bandwidth, theworsening of SFDR is just 2/3 * X(dB). The attention will befocussed on Vertical Cavity Surface Emitting Laser (VCSEL),which is the last developed after Fabry-Perot (FP) andDistributed Feedback (DFB). Typical SFDR range of the DFBlasers is 90-110 dB.Hz^2/3. Both FP and VCSEL lasers have lowerSFDRs than DFB, this is mainly due to the multimodebehaviour. Before discussing the prior art a briefintroduction to the most popular solid state lasertechnologies and some relevant applications are given inorder to introduce the technical problem.
• FP lasers are edge-emitting devices; that is theemitted light pencil comes out parallel to the substrate atone side of an active resonating cavity placed in the middleof a wafer structure. Top and bottom layers of the waferconsist of two slabs of semiconductor material speciallydesigned to behave as two faced mirrors reflecting the lightback and forth the solid state cavity. Electric current flowsthrough the device from the top slab to the bottom forstimulating the emission of light inside the active layer.The emitted light is spatially and temporally coherent butnot entirely monochromatic because more wavelengths areemitted inside a narrow window around a central wavelength.This characteristic of the emitted light is said "multimode".FP lasers at the second optical window coupled to eithermonomode or multimode optical fibres are commonly used indigital communications such as Local Area Networks (LAN). TheFP lasers have some limitations because of their multimode characteristic, in particular:
• due to the coupling between multimode emitted light anddispersive optical fibres, the shape of the transmittedsignal is distorted as it travels along the fibre; inconsequence of that the maximal length of the fibre haulshall be reduced in comparison with monomode;
• Wavelength Division Multiplexing (WDM) inside the singleoptical fibre is prevented because an extremely chromaticpurity of the light is needed. Multimode fibres suitablefor WDM are doped for obtaining adjacent optical windowsof minimum losses.
• 1. The opportunity of using FP lasers for transmitting analogband-pass optical signals is disclosed in the article:"Dynamic Range Requirements for Microcellular PersonalCommunication Systems Using Analog Fiber-Optic Links",authors: J.C. Fan, C.L. Lu, and L.G. Kazovsky, publishedin IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES,VOL. 45, NO. 8, AUGUST 1997. The object of this IEEEarticle is that to study the quality requirements of thelaser to be used in a centralised optical fibreinfrastructure for a wireless Personal CommunicationSystem (PCS). PCS is a modern cellular telephony systemcharacterised by cells of reduced dimensions, i.e.:100-500 m radius, as indicated by WACS/PACS BellcoreStandard TR-INS-001313,1994. For this aim a comprehensivewireless/optical model is provided to perform computersimulations. The hypothetical simulated scenario foreseesa centralised Base Station optically connected to manyremote antennas serving as many microcells. Analog signalsin the PCS band around 2 GHz (Rx 1850-1910 MHz ; Tx1930-1990 MHz) cross the fibres. The only uplinkconnection has been studied because this connection is themore penalised one (BTS antenna collects a larger interference than the mobile). Channel band of 270 kHz andQPSK modulation are assumed. The uplink channel includesan AGC which keeps the modulation depth of the laserconstantly near to the optimum. Fast Power Control able tocounterfeit the Rayleigh fading and the log-normalattenuation is introduced in the model. Typical penalty ofthe wireless system, such as: Rayleigh fading, cochannelinterferences, adjacent channel interferences, etc.,together with laser penalty, such as noise andintermodulation are considered. A mathematical expression(5) including laser IMD3 product (coefficientb₃) has beendeveloped for calculating the whole PCS+Laser SIR (Signalto Interference Ratio). Coefficientb₃ is separatelyresolved by SFDR and the result introduced in the originalexpression (5) in order to evaluate the impact of the SFDRon the whole SIR. Plotting are given on the PCSavailability as function of the SFDR parameter of thelaser. The conclusion is that (due to the statisticaluncertainty of the wireless losses and interferences) anSFDR value in the range 72-83 dB.Hz^2/3 is acceptable andeither single-mode DFB or multimode FP semiconductorlasers can be used. Nothing is said about the use ofVCSELs. Besides, given the theoretical nature of thearticle, the suitability of FP lasers is taken from apaper titled: "Interrelationship of wireless services withoptical networks", presented by K. Emura at the LEOS'95Conf., Sanf Francisco, CA, October 1995.
• DFBs are monomode edge-emitting lasers having higherwavelength resolution than the previous FPs, so as toovercome the underlined multimode limitations. A DFB laser issubstantially obtained from the FP structure by inserting agrating layer into the active resonant cavity. The gratinglayer acts on the light reflection in a way that the final effect is the selection of a single wavelength into theactive cavity, this means that the emitted light is monomode.Due to their inherent manufacturing complexity DFB lasers aremuch more expensive than FPs, even of 1000 times, but theincreased costs could be absorbed by the suitability to beemployed with larger WDM spectral allocation. DFB lasersemitting in third optical window coupled to single-modeoptical fibres are used in long-haul optical rings.
• Thomson-CSF and Miteq have jointly developed a monomodeDFB Analogue Microwave Optical link. High speed photodiodeand the E-2000™ optical connector system from Diamond isused in the link. The relative brochure is titled: "MDDMicrowave optical link". Many applications are foreseen,i.e.: remoting, timing and reference signal distribution,telemetry, measurement, etc. Various possible modulations aredepending on the particular applications. 12 GHz bandwidthand low noise are available for analogue applications. Theproposed optical link is very expensive, due to the high costof the DFB laser and strict mechanical tolerances of theDiamond connector, so that massive applications arediscouraged. Besides the manufacturing costs, also theinstallation costs have to be considered; in fact theoperator shall perform in-site difficult heading and checks.
• Differently from the preceding lasers FP and DFB,multimode VCSEL lasers are vertical-emitting devices becausethe light pencil is emitted perpendicularly to the substratethrough the semitransparent top slab electrode of the waferstructure. Top and bottom electrodes which delimit the activeresonant cavity are DBR (Distributed Bragg Reflector)structures made of a stack of alternate mirrors. The lightpencil is narrower and more circular than the preceding FPsand DFBs and the coupling to the optical fibres is madeeasier consequently. VCSEL lasers are about a third less expensive than FPs because, thanks to the emission throughthe substrate, they can be tested directly on themanufacturing substrate before cutting. Multimode VCSELdevices emit preferentially inside the first optical window(850 nm). Being the embodiment of the invention in subjectbased on VCSEL, the advantages of this technology are listedin the following:
• Circular symmetric far field with a small angle ofdivergence, so that the emitted beam can be injected verysimply and efficiently into optical fibres.
• Potentially low cost, because the technologies used forVCSEL are the same as those used for the fabrication ofintegrated electronic circuits.
• Low threshold currents and efficiencies of over 50%.
• No light reflected back towards the laser re-enters theresonator, thanks to the DBR (Distributed Bragg Reflector)output coupler. As a consequence, they show a typicalvalue of Relative Intensity Noise (RIN) of -125 dB/Hz,which is lower than the typical edge-emitter values.
• MTBF (Mean Time Before Failure) exceeds ten million hoursat room temperature and moderate currents.
• The primary area of application of low power VCSELs isshort-haul gigabit Ethernet in baseband digital communicationthrough multimode fibres. In this area they cover near 95% ofthe applications. The low cost and low power supplyrequirements of VCSELs make them attractive candidates forthose applications, such as antenna remoting, that requiremany optical links. VCSELs digital optical links arecurrently used for antenna remoting, the Applicant SIEMENS isactive in this field. The primary aim of antenna remoting isthe reduction of RF transmission power losses due to thelength of the coaxial cable connecting the indoor transmitter to the tower antenna. Thanks to the digital optical link, thebaseband processors are kept at the base of the tower, whilethe remaining transmitting and receiving parts are placeddirectly on the rear of the antenna. The same digitalimplementation is used in complex smart antenna systems whichforesee a certain number of SWR (Software Radio) chains(typically 4 or 8) each connected to a respective antennaelement of a directive array.
• The suitability of VCSEL analog optical links forantenna remoting (limited to the PCS system) is discussed inthe article titled: "Multimode VCSEL links for RFCommunication", authors: Harry Lee and Rajeev J. Ram, SponsorU.S. Navy - Office of Naval Research; published in RLEProgress Report Number 141, pag.160-161. Thanks to the analogoptical link the outdoor apparatuses can be furthersimplified. In particular, the following elements areshiftable indoor: both D/A and Up frequency converter fordownlink; for uplink both A/D and Down frequency converter.In the Harry Lee's article two Figures are given forrepresenting contours of constant SFDR values measured at theend of two optical links 27 m and 150 m long, as function ofboth bias current and frequency. The shortest link shows anSFDR peak of 96 dB.Hz^2/3 (1 Hz BW) nearly constant inside aband of 200-300 MHz centred around 3.8 GHz, for optimum biascurrent of 20 mA. The longer 150 m link shows a maximum SFDRof 91 dB.Hz^2/3 nearly constant inside a band of 100-200 MHzcentred around 2 GHz, for a bias current greater than 20 mA.
APPROACH TO THE TECHNICAL PROBLEM
• A sound technical problem concerning VCSEL analogoptical links is that to extend the antenna remoting to themost popular cellular systems spanning wider distances thanthe ones typical of PCS. To this aim the SFDR parameter ofthe whole link must be investigated. A partial teaching in this direction is given in the cited article of Harry Lee andRajeev J. Ram. The highest SFDR value of 91 dB.Hz^2/3 measuredwith 150 m link against the needed PCS range of 72-83 dB.Hz^2/3makes it suitable for PCS. Nevertheless this suitability tobuild up the PCS's fibre infrastructure should be verified atthe maximal length. The shortest maximal length of a PCSoptical fibre connection is at least 500 m and occurs incorrespondence of a centralised base station (as shown infig.1). By a simple extrapolation of the difference[SFDR_{27 m} - SFDR_{150 m}] measured by Harry Lee to the difference[SFDR_{150 m} - SFDR_{500 m}], the final SFDR value of the analogVCSEL optical link should be reduced to 71.6 dB.Hz^2/3 (whichis lightly under the minimum PCS value of 72 dB.Hz^2/3).
• Longer optical links than PCS are needed to implement acentral fibre infrastructure inside GSM or UMTS clusters.Urban cells with radius greater than 1 km are usual in GSMand the same should be for UMTS. Rural cells, as a rule, havegreater radius than urban cells. Simulations to set suitableSFDR ranges for GSM and UMTS are presently unknown,nonetheless SFDR values of 72-83 dB.Hz^2/3 are largelyinadequate in Software Radio applications where programmablewideband transceivers with high linearity are needed.Unfortunately, when attempting to increase the length of theanalog VCSEL optical link its SFDR is reduced much more thanmonomode, also the channel bandwidth is reduced. The givenexplanation is that for longer multimode links the modaldispersion dominates the link performance and the dynamicrange follows the frequency response of the optical fibre,which depends on how the mode groups are excited in thefibre, which in its turn depends on the bias current of theVCSEL setting the gain suitable for a given fibre length.This complex dependence introduces large SFDR variations withthe length of the fibre and also sharp nulls for certain frequencies, as shown by the comparison of the two Figuresreported in the article of Harry Lee et al., were the greatvariation on the shapes of constant SFDR contours passingfrom 27 to 150 m link is put in foreground. The complicatedlaw which the SFDR parameter obeys to, makes difficult tomatch the required quality of the VCSEL analog optical linkwhen the fibre lengths are involved in the practical casessuch as: PCS, GSM, and UMTS. This drawback strongly limitingthe useful length of the VCSEL analog links. In any case themaximum measured SFDR of 96 dB relevant to 27 m optical linkis slightly under the limit of 97 dB required for CableTelevision (CATV).
• Known analog multimode optical links shall include aminimum number of elements indispensable for the opticaltransmission, i.e.: a RF preamplifier as far as possiblelinear, the laser diode, a matching network between the two,a multimode optical fibre, a PIN diode followed by a RFamplifier and another matching network between the two. Theknown designs for the optical receiver, which includes thePIN diode and the RF amplifier, are substantially two: thefirst is calledhigh impedance orintegrating front end, thesecond is calledtransimpedance amplifier. Thehigh impedanceapproach provides the means for achieving minimum circuitnoise. However this approach usually results in reduceddynamic range and the possible requirements of individuallyequalised amplifier. Thetransimpedance amplifier approach isfinding widespread use because it: a) is capable of widebandwidths, b) provides greater dynamic range, and c) hasnoise performance which can approach that of the firstapproach. Detailed discussions of these two approaches arefound in the literature.
• As far as concerns VCSELs, many Honeywell ApplicationSheets give exhaustive explanations on the intrinsic noise and distortions. A relevant parameter of VCSEL is theelectro-optical conversion gain function P(I), where P is theemitted light power and I is the forward excitation currentthrough the VCSEL. This function is quite linear above athreshold current I_TH. The slope efficiency η = ΔP/ΔI isstrong dependent on the temperature. Obviously the lineartrend of the slope cannot be maintained indefinitely andsaturation happens sooner or later, in such a casedistortions like Amplitude Modulation to Amplitude Modulationconversion (AM/AM) occurs. Variations on the bias current ofVCSEL affect the intrinsic intermodal dispersion inside theoptical fibre (see Harry Lee et al.) deviating the channelcharacteristics from the ideal. In this case AM/PM distortionoccurs. Deviations from the ideal also occur on the detectedsignal by effect of the gain compression of the various RFamplifiers at the highest power levels. In such a case bothAM/AM and Amplitude Modulation to Phase Modulation conversion(AM/PM) are possible. Those distortions are particularrelevant in case a final RF power amplifier is useddownstream the optical receiver to feed a transmittingantenna. In this case the AM/PM distortion of the final RFpower amplifier gives an additional contribution to the AM/PMdistortion due to the intermodal dispersion inside theoptical fibre. Whether said deviations from the linearity arenot adequately counteracted, the signal at the end of theoptical link is no more a faithful copy of the input and theoverall quality of the link is worsen. Moreover noisenegatively affects the fidelity of the optical link and shallbe kept as low as possible. The high SFDR values requested bysome actual standards force the designers to minimise thetolerated noise; that is a sound design should have care ofreducing distortions together with noise. Another parameterto be considered in a sound design is the possible peak-to-average power ratio of the input signal, because onthe high peaks all the active devices might be saturated andintroduce distortions, consequently. It may happen that uponthe peaks the input preamplifier is not more linear, so asthe VCSEL and the RF power amplifier, and the losses of theoptical fibre increase due to a side change of the intermodaldispersion. Some countermeasures are known to prevent theseside effects, i.e.: the impedance matching between VCSEL andits driving amplifier; the impedance matching between the PINdiode and its downstream amplifier; the closed-loop controlof the VCSEL bias current; the gain and phase linearisationof the RF final power amplifier.
OBJECTS OF THE INVENTION
• The object of the present invention is that to overcomethe drawbacks of the prior art and to indicate an analogueVCSEL optical link showing acceptable SFDR values compatiblewith the services involved as the length of the fibreincreases.
• A first preferably length being at least 150 m to matchup to typical lengths of the intra-building connectionsinvolved in CATV service. This length also exceeds thetypical lengths suitable to connect indoor apparatuses withtower antenna, in particular an antenna of a line-of-sightradio relay, either of the point-to-point orpoint-to-multipoint type, or an array of antennas suitablefor beamforming.
• A second preferably length being at least 500 m to matchup to a small-size fibre infrastructure suitable forconnecting a centralised base station of a pico/microcellularsystem to a pool of remote antennas serving as many cells.
• A third preferably length being at least 1.6 km to matchup to a fibre infrastructure suitable for connecting acentralised base station of a second or third generation cellular telephone system (GSM or UMTS) to a pool of remoteantennas serving as many urban/rural cells.
• Any other lengths greater than 1.6 km are preferred tomatch up to wider cells on condition that the quality of theoptical link is maintained. A channel bandwidth of 5 MHz isdesirable with all the above lengths in order to accommodateat least one UMTS channel (WCDMA, UTRAN-TDD-HCR). A channelbandwidth of 6 MHz is desirable to accommodate at least oneCATV channel (as per the U.S. TV Frequency Standards) and8 MHz in the European countries.
SUMMARY AND ADVANTAGES OF THE INVENTION
• To achieve said objects, an analog VCSEL optical link isdisclosed in the main claim of the present application,relatively to a minimal configuration including the followingelements: a first RF amplifier receiving an analog RF signaland generating a current thereof to drive a VCSEL lasercoupled to a multimode optical fibre of a given length; afirst matching network between the VCSEL and the first RFamplifier; a PIN diode followed by a second RF amplifier anda second matching network between the two; an RF signalenvelope detector generating a control signal for a variablephase shifter placed downstream the second RF amplifier tocompensate for the AM/PM distortion introduced by the opticalfibre. In order to further characterise the foregoingcombination: the first RF amplifier is a linear preamplifier;the first impedance matching network is of the wideband type;the second RF amplifier is a low-noise (FET) transimpedanceamplifier; and the second impedance matching network isdesigned to minimise the noise figure of the transimpedanceamplifier.
• The compensation of the optical dispersion driven by theamplitude envelope is surprising indeed for the followingreasons: a) the intermodal dispersion is an unintentional optical effect, and b) the complex relation existing betweenthe amplitude envelope and the AM/PM distortion inside anoptical fibre is yet not completely cleared up. Furthermore,the PIN diode and the transimpedance amplifier with theinterposed circuitry, is a new approach in designing opticalreceiver having the advantage of minimum noise of thehighimpedance approach together with greater dynamic range andwide bandwidth of thetransimpedance amplifier approach.
• The minimum configuration of the optical link ischaracterised by low amplifications, nonetheless performancesobtained with short links are comparable to the performancesof optical links based on very expensive lasers and monomodeoptical fibres. As an immediate consequence, the proposedanalog VCSEL optical link is the cheapest solution forintra-building connections in CATV applications. Besides, itallows significant cost saving in those base stations wherelarge optical link deployment is needed, such as the onesusing smart antennas for reception beamforming.
• A first alternative embodiment of the invention includesa third RF amplifier placed immediately before the variablephase shifter. The two RF amplifiers placed in tandemdownstream the PIN diode provide additional amplification toallow further applications of the invention, such as tobuild-up a fibre infrastructure between a centralised BaseStation of a cellular telephone system and a pool of remoteantennas used for downlink transmissions. In this case thethird RF amplifier is a final power amplifier connected to atransmitting antenna, and the variable phase shifter alsocompensates for the AM/PM distortion introduced by the RFpower amplifier on the power peaks.
• A second alternative embodiment of the invention,depending on the first one, includes a gain linearisationnetwork designed to compensate for the AM/AM distortion of the final RF power amplifier on the power peaks of RFmodulated signals characterised by high peak-to-average powerratios.
• A third alternative embodiment of the invention,independent of the preceding two, includes a gainlinearisation network designed to compensate for the AM/AMdistortion of the first RF power amplifier due to thefollowing combination: a) the high amplification which isneeded to drive the VCSEL adequately for the longest opticallinks; b) the presence of RF modulated signals characterisedby high peak-to-average power ratios.
• The combined teaching of the invention and itsalternatives allows to improve noticeably the SFDR of theanalog VCSEL optical links, making possible to prolong theirlength well beyond the indicated distance of 150 m,consequently. Furthermore, a great variety of RF analogsignals can be transmitted/received through the link of theinvention, for example, spacing from a pure sinusoid tomodulated signals having a bandpass spectrum up to 100 MHzbandwidth, freely allocable into a wide frequency rangespanning from about 200 MHz up to the highest microwavefrequencies of VCSEL (few GHz). There are not limitations onthe type of modulations, i.e.: AM, FM, PM, GMSK, CPM, N-PSK,N-QAM, etc. Single carrier or multicarrier modulated signalsare transmissible through the said optical link arranged forpoint-to-point or point-to-multipoint communications. It canbe appreciated that in case of multicarrier transmissionswith variable number of carriers and/or high-index QAMmodulations, the present invention is able to operate inpresence of great peak-to-average power ratios.
• Because of good performances are held when the fibrelength increases other than 1 kilometre, a preferredembodiment of the invention surely includes the first alternative embodiment, but most advantageously might includeeither the second or the third alternative embodiment, orboth, without limitation to the type of cellular system ofthe second (GSM) or third generation (UMTS).
BRIEF DESCRIPTION OF THE DRAWINGS
• The features of the present invention which areconsidered to be novel are set forth with particularity inthe appended claims. The invention, together with furtherobjects and advantages thereof, may be understood withreference to the following detailed description of anembodiment thereof taken in conjunction with the accompanyingdrawings given for purely non-limiting explanatory purposesand wherein:
• fig.1 shows a cellular system architecture based on acentralised base station connected by means of couples ofoptical fibres (one fibre for each direction oftransmission) to as many remote antennas located insiderespective cells/microcells; the optical links includingthe teaching of the present invention;
• fig.2 shows a CATV system whose intra-building opticallinks include the teaching of the present invention.
• fig.3 shows the transmitter and receiver parts at the twosides of a couple of optical fibres of fig.1;
• fig.4 shows the transmitter and receiver parts of thepreceding Figure implemented in accordance with thepresent invention;
• fig.5 shows a generic Device Under Test (DUT) tested forits 3rd order Intermodulation Distortion (IMD3);
• Figures 6 and7 schematise IMD3 distortion and IP3concept, respectively;
• fig.8 shows the DUT of fig.5 including a VCSEL opticallink of the known art;
• Figures 9a and9b show two plots of the gain vs. frequency measured for the DUT optical link of fig.8,with 500 m and 1.6 k m fibre lengths, respectively;
• Figures 10 and11 show circuital details of two networksLASINT and PHOTINT belonging to the optical links offig.4;
• Figures 12a and12b show a plot of the gain vs.frequency relevant to the DUT of fig.8 upgraded with thetwo networks LASINT and PHOTINT of fig.10 and fig.11 with500 m fibre length;
• Figures 13a and13b are similar to Figures 12a and 12bwith 1.6 k m fibre length;
• fig.14 shows circuital detail of a matching network PHCinternal to optical links of fig.4;
• Figures 15a and15b show a CDMA output spectrum and thecorresponding ACPR interference relevant to the DUT offig.8 upgraded with the three networks LASINT, PHOTINTand PHC of Figures 10, 11 and 14, for 500 m fibre length;and
• Figures 16a and16b are equivalent to Figures 15a and 15bwith 1.6 k m fibre length.
DETAILED DESCRIPTION OF AN EMBODIMENT OF THE INVENTION
• With reference tofig.1 a Base Station BS is visible atthe centre of a cluster including cells and/or microcellsbelonging to a cellular telephone system. A remote antenna isinstalled on a build placed at the centre of each cell of thecluster. Remote antennas and the centralised base station areconnected to each other by means of couples of multimodeoptical fibres, one fibre for each direction of transmissionin case of FDD duplexing (only 1 fibre for TDD half-duplex).Conventionally, the transmission from the base station BS toa remote antenna is said downlink (or direct link), while theopposite direction is said uplink (or reverse link). Morehand-held mobile phones, either Mobile Station MS or User Equipment UE, are connected via-radio with the respectiveremote antenna of the serving cell. The centralised basestation is connected to a Base Station Controller (BSC), thelatter is connected to other base stations, to a PublicSwitch Telephony Network (PSTN), and to an Internet NetworkIP. The depicted cluster might belong to any of the mostpopular cellular systems, for example: GSM-GPRS 880-960 MHz;DCS-GPRS 1710-1880 MHz; PCS 1850-1910 MHz; and the followingUMTS standards located in the proximity of 2 GHz: UTRA-FDD3.84 Mcps; UTRA-TDD-HCR 3.84 Mcps; UTRA-TDD-LCR: 1.28 Mcps;TD-SCDMA 1.28 Mcps; etc.). Considering a Root Raised Cosine(RRC) filter response with a roll-off α=0.22 and a bandwidthequal to the chip rate, it results a channel spacing of 5 MHzfor 3.84 Mcps chip rate, and 1.6 MHz for 1.28 Mcps. Dependingon the system represented in fig.1, cells from 100 m up to1.6 km radius, and more, are possible. Each singular opticalfibre of the infrastructure represented in fig.1 is crossedby an optical signal which optically reproduces an analogradiofrequency signal either to be transmitted by, orreceived at, the relevant antenna. Type of modulation, code,and burst format of the optical signal are the ones asspecified at the on-air interface of the represented system;for example: the GSM Um interface, the UMTS Uu interface.,etc. The interface between base station BS and base stationcontroller BSC is the one as specified by the relevantstandard. The base station controller BSC is connected to thePSTN and IP networks through the A interface. Theconfiguration with the base station BS at the centre of thecluster is the one having the highest symmetry and theminimum length of the maximal fibre haul. It can beappreciate that the centralised architecture save a lot ofbase stations together with the respective tower antennas.The fibre infrastructure allows each subscriber station MS/UE to communicate with the centralised base station BS either inoriginating or terminating calls. Base station controller BSCroutes calls originated inside a cluster relevant to acentralised base station BS either towards other centralisedbase stations BS or the PSTN / IP network, and towards therelevant base station BS calls to be terminated inside therelative cluster.
• A variant of the cluster of fig.1 (not shown forsimplicity) is referred to a smart antenna system including afibre infrastructure similar to the one visible in fig.1,except for one or more cells including an M-elements antennaarray each connected to the base station BS by means of 2Manalog RF optical links, in case of FDD duplexing. As known,an antenna array together with the connected opticalreceivers and transmitters allow to perform beamforming,either 360° or tri-sectorial, for the aim of interferencereduction and spectral efficiency increasing.
• Fig.2 shows a CATV system using optical fibres forintra-building and extra-building connections. With referenceto the Figure, a CATV centre includes a ANALOG MUX (videoserver) which connects an analog film archive (tapes orcassettes) to a plurality of optical transmitters TRXs.Analog signals at the outputs of TRX transmitters areconveyed on an analog optical cable FB directed towards abuilding in the neighbouring. A computer PC is connected tothe video server for scheduling the pay-per-view and/orsubscription services, and to a modem MD for collecting theservice requests from subscribers. A standard telephone-setis further connected to a twisted pair TP for the same aim.The optical cable FB reaches a signal distributor DIST fromwhich singular multimode optical fibres depart towardsrespective flats. Inside each flat the incoming fibre isdirected to an analog set-top box STB connected to a TV screen. In a large area there are buildings quite far to theCATV centre, this prevents the use of a multimode opticalcable FB because of the excessive losses. A monomode FB cableis profitably used to transmit an FDM optical signal carryingmany analog TV channels. At the other end of the FB opticalcable a regeneration of the optical signal inside thedistributor DIST could be preferable for many reasons. Forthis purpose the signal distributor DIST includes an opticalreceiver for each optical fibre of the optical cable FB, andas many analog optical transmitters as the departing fibrestowards the set-top boxes STB on the flats. Each set-top boxSTB includes an analog optical receiver. Regeneration alsoimplies optical-electrical-optical conversion. Afteroptical-to-electrical conversion the various RF signals areamplified, if that is the case frequency translated, androuted to the various optical transmitters. VCSEL lasers andmultimode optical fibres are preferred forelectrical-to-optical conversion and intra-building opticaldistribution, due to their low cost and the small distances(lower than 150 m). Optical transmitters and receivers at thetwo ends of each local link are similar to the ones that willbe disclosed successively.
• With reference tofig.3 the Base Station BS includes thefollowing elements: as many analog optical transmitters BSOTXas the departing multimode fibres DL towards the remoteantennas, as many optical receivers BSORX as the incomingmultimode fibres UL from the remote antennas, a basebandprocessor, and a Remote antenna controller. The basebandprocessor is connected to all the transmitters BSOTX and thereceivers BSORX, and also to the Remote antenna controllerwhich in this particular embodiment takes the place of thebase station controller BSC of fig.1. The remote antennaincludes: one optical receiver RAORX, one optical transmitter RAOTX, a duplexer filter and the external antenna. Acirculator replaces duplexer filter in TDD systems.
• Fig.4 shows BSOTX, BSORX, RAORX, RAOTX apparatuses witha first level of detail. The two BSOTX and RAORX blocksconnected at the two ends of the multimode fibre DL includethe optical link object of the present invention; the link isused to feed the transmitting antenna with the RF signal. Inthe opposite direction the RAOTX and BSORX blocks connectedat the two ends of the multimode fibre UL include the sameoptical link used for transferring the uplink signal receivedat the antenna towards the base station BS. While the twooptical links are near the same, the two transmitters BSOTXand RAOTX are different to each other, so as the tworeceivers RAORX and BSORX. The transmitter BSOTX includes thefollowing cascaded elements: a baseband transmissionprocessor TX PROC belonging to the BASEBAND PROCESSOR offig.3, a modulator MOD, a D/A converter (not shown forbrevity), a frequency up-converter UPC, a linear RF amplifierRFA, an impedance matching network LASINT, and a multimodeVCSEL laser. In operation, the processor TX PROC performstypical baseband operations on the incoming routed signal inorder to provide I-Q digital signals to the modulator MOD.The modulated signal is digital-to-analog and IF converted,and forwarded to the up-converter UPC which also receives alocal oscillator signal LO and generates the analog RF signalin the assigned channel band. The RF signal is amplified bythe RF amplifier RFA and forwarded to the impedance matchingnetwork LASINT which provides optimal interfacing with theelectrical input of a multimode VCSEL laser. The latterconverts the modulated RF signal directly in an opticallymodulated light beam feeding the multimode fibre DL.
• The receiver RAORX includes the following cascadedelements: a diode PIN (photodetector), a network PHOTINT, an RF power amplifier PWA, and a phase distortion compensationnetwork PHC. In operation, the PIN diode is optically coupledto the fibre DL for converting into electrical the modulatedoptical signal. The converted signal is again the modulatedRF signal at the input of VCSEL, plus noise andinterferences, and the intermodulation distortion generatedinside the laser, plus the overall effect of the intermodalcoupling along the optical fibre DL. The resulting RF signalis forwarded to the network PHOTINT which provides optimaldetection and pre-amplification at the output of PIN diode.The RF signal at the output of PHOTINT is forwarded to theinput of the RF power amplifier PWA. The amplified RF signalat the output of PWA is forwarded to the phase distortioncompensation network PHC, which attempts to cancel the AM/PMdistortion before irradiating the RF signal. Additionalimprovement of the optical link performances can be achievedby introducing a gain linearisation network.
• The transmitter RAOTX includes the following cascadedelements: a low noise amplifier LNA, a matching network LSINTand a VCSEL laser. These blocks operate as the correspondingblocks RFA, LSINT and VCSEL included in the transmitterBSOTX.
• The receiver BSORX includes the following cascadedelements: a photodetector diode PIN, a network PHOTINT, an RFpower amplifier RFA, a phase distortion compensation networkPHC, a downlink frequency converter DWC, an A/D converter(not shown for brevity), a demodulator DEM, and a basebandreception processor RX PROC belonging to the BASEBANDPROCESSOR of fig.3. In operation, the first four blocksoperate as the equivalent blocks included in the receiverRAORX, except for lower amplification performed by RFA inrespect of PWA. The mixer DWC receives both the RF and thelocal oscillator signal LO (LO is unique for TDD) and converts the RF signal to IF. The IF signal is down convertedand demodulated, and the baseband signal is forwarded to theblock RX-PROC. The latter performs typical basebandoperations reciprocal to the ones performed by the blockTX-PROC, obtaining a baseband digital signal to be routedtowards the final destination.
• Given the substantial equivalence of the two opticallinks used for uplink and downlink, an uniform denominationof the amplifiers is used; precisely: RFA and LNA are namedA1 while PWA and RFA are named A2. The uniform denominationallows to study the two optical links as an unique two-portDUT (Device Under Test) like the one visible in theFigures 5and8.
• With reference tofig.5 the DUT is considered as ageneric two-port network affected by noise andnon-linearities. At the input the thermal noise (-174 dBm/Hzat 290° K) is always present. The DUT is characterised byGainG [dB], phase offset Φ[rad/degrees] between input andoutput signal, Noise FigureNF [dB], andSFDR [dB.Hz^2/3]. Atthe input of the DUT a signal x(t) is present which issupposed to be narrowband around a frequency fc, so that itcan be represented by means of its baseband complex envelope:x(t) =A(t)ej(t)whereA(t) and (t) are the amplitude and the phase of signalx(t), respectively. At the output of the DUT, supposed to bebandpass and memoryless, a signaly(t) is present, whosebaseband complex envelope is:y(t) =G(A(t))ej{(t)+Φ(A(t))}.The two functionsG(A(t)) and Φ(A(t)) denote the overall AM/AMdistortion of the gain functionG and the overall AM/PMdistortion of the phase function Φ, respectively. Expression (2) shows that both gainG and phase Φ are depending on theamplitudeA(t) of the input signalx(t). Especially with M-QAMmodulation characterised by high-M and in presence ofmulticarrier modulated signals with a variable number ofcarriers, the amplitudeA(t) is subjected to highpeak-versus-medium values; in these circumstances both AM/AMand AM/PM distortions might be considerable and the relevantside effects shall be counterfeited. FunctionsG(A(t)) andΦ(A(t)) can be approximated by their points experimentallydetermined by means of a test signalx(t). Once the twodistortion curves have been approximated, some compensatingnetworks are synthesised. To meet this goal, in accordancewith the teaching of the invention the envelope ofy(t) isdetected by the relevant linearisation/compensation networksused in the optical link. An alternative way to determine theAM/AM and AM/PM curves is that of developingy(t) into alimited power series and equatingG and Φ to the elements ofthe series. DUT of fig.5 is submitted to the classicaltwo-tone test just for measuring the relevant 3rd orderIntermodulation Distortion (IMD3). In this case the testsignalx(t) at the input of DUT is the summation of twosinusoidal tones of equal level and frequency F1 and F2,closely spaced to each other, whose frequency offset dependson the channel bandwidth of the effective signal.
• With reference tofig.6 at the output of the nonperfectly linear DUT of fig.5 the following elements will bepresent:
• two signals at frequency F1 and F2, and equal levelP1 = P2 in correspondence of the two tones;
• third order intermodulation products P3 and P4, of equallevel, at frequency F3 = (2F1-F2) and F4 = (2F2-F1); frequencies F1 and F2 can be selected in a way thatfrequencies F3 and F4 fall in the channel band;
• the noiseN [dBm/Hz] = -174 +G +NF.
• The spectrum visible in fig.6 suggests that among allthe spurious tones generated by a non perfectly linear devicecrossed by a bandpass signal, there are someone very closelyspaced to the original tones which represent potentialadjacent channel spurious signals. Designers shall have ameasure of the grade of linearity of the used electronicdevices in order to evaluate the interference spectrum,therefore relevant parameters such as SFDR and InterceptPoint (usually 3rd order IP3) have been introduced. In fig.6theSFDR (Spurious Free Dynamic Range) evaluated at theoutput of the DUT isP [dB] =P1-P3 in correspondence ofwhich P3 =N (noise). Being P3 a power level [dBm] whileN isa spectral density [dBm/Hz], it be understood thatN ismeasured into 1 Hz bandwidth.
• The concept of Intercept Point (IP) depicted in fig.7was introduced some years ago and has become a useful andpopular specification for comparing the quality of variousnon-linear electronic components. The intercept point is thetheoretical level at which on a plot of output vs. inputpower levels, the two-tone distortion products transfer curveintersects the single tone transfer curve. Intercept is basedon an extrapolation from measures at low power and assumesideal slopes for the output tone level dependencies. Theideal slopes of the two-tone signal is three time the slopeof the single tone (3 dB against 1 dB increase in alogarithmic representation). The following expression isvalid for 3rd order Intercept Point (IP3in) evaluated at theinput of the DUT:IP3in [dBm] =P1 +P/2 -G. LetD [dBm] =IP3in - NF be the Dynamic Range Number, thefollowing relation is true:SFDR = (2/3) x (174 +D). The two parametersIP3in andSFDR are indicated in the abscissa.
• Fig.8 shows a simplified embodiment of the DUT of fig.5for an analog VCSEL optical link. This link includes twoamplifiers A1 and A2, other than the optical VCSEL and PINdevices and the multimode optical fibre. The aim of this DUTis that to provide a reliable comparison for theperformances of the invention. According to this aim theVCSEL, the multimode optical fibre, and PIN diode are thesame used by the invention. The used VCSEL and PIN diode arecommercial low cost devices, i.e.: Honeywhell VCSELHFE419X-521 and Honeywhell PIN HFD3180-102. Three insertionpointsa, b, c are indicated by arrows along the link incorrespondence of successive insertions of the networksLASINT, PHOTINT and PHC indicated in fig.4. Each insertionpoint represents a gradual improvement of the optical linkobtained by the introduction of a corresponding network, inaccordance with the present invention. The embodiment offig.8 should be further simplified to consider the minimalconfiguration of the link; in this case the amplifier A2 isthe one included in the network PHOTINT. Test tones P1 and P2are spaced 10 MHz apart around a central frequency taken formere convenience at 1685 MHz. In such a case it results:F1 = 1680 MHz, F2 = 1690 MHz, F3 = 1670 MHz, andF4 = 1700 MHz. The power at pointb (fig.8) is noticeablylower than the power transmitted at pointa because of theinsertion losses of the multimode optical fibre. Afundamental parameter of the invention in subject is themaximum length of the optical fibre. This length isdetermined from the criterion that the signal power reachingthe receiver (PIN plus A2) must exceed the receiversensitivity. That is: 'PT -L(lmax)≥PR where:
• P_T is the end-of-life transmitter power (in dBm) whichcan be injected into the fibre;
• P_R is the worst-case receiver sensitivity (in dBm),including the side effects associated with the dispersiveproperties of the medium (i.e. Intersymbol Interference(ISI)), and
• L(1) is the insertion loss in dB of an optical fibre oflength1. As far asL(1) is concerned, two losses are tobe distinguished: a first one is the light attenuation,typically 1 dB/km, the second one is the attenuation ofthe RF signal which modulates the light. This second lossis considerably greater than the first although thereason is not presently well understood; in our case thisloss is near 5 dB with 500 m fibre.
• Expression (3) is useful to set a rough value of thenominal excitation current of VCSEL, taking in dueconsideration the threshold current I_TH and thepeak-to-average power ratio of the RF signal. Due to theuncertainty of the insertion lossL(1) the rough value shallbe experimentally corrected.
• The followingFigures 9a and9b show two plots of thegainG vs. frequency measured for the VCSEL optical link offig.8 with 500 m and 1.6 km optical fibres, respectively.Only a limited frequency range is reported, for brevity. Thetwo plots show flat gain in the depicted frequency range; theflat characteristic is held well outside the visible range.The measured value of the input dynamic rangeD with500 mfibre is-43.5 dBm, corresponding to anSFDR of87 dB.Hz^2/3.With1.6 km fibre the values becomes D =-57 dBm,corresponding to anSFDR of78 dB.Hz^2/3. The central frequencyof the two-tones is moved from 200 MHz to 2.5 GHz withoutappreciable variations of the indicated values.
• Fig.10 shows the network LASINT of fig.4 to be insertedin the optical link at pointa of fig.8. The network includesa discrete decoupling capacitor C1 of 2.2 pF seriallyconnected to a transmission line L1. The two ends of theseries of C1 and L1 correspond to the input and the output ofthe LASINT network. Externally to LASINT the VCSEL laser isserially connected to a pull-up resistance R1 towards thepolarisation voltage -VL. Resistance R1 is dimensioned todrive the VCSEL optimally. Line L1 is a microstrip on FR4dielectric substrate (ε_r = 4.7) 0.3 mm thick, dimensioned tomatch, together with capacitor C1, the input impedance ofVCSEL to the output impedance of the amplifier A1 (fig.8) andavoiding reflections. Dimensions of line L1 are thefollowing: width w = 0.2 mm,length 1 = 12 mm. The dimensionsof microstrip L1 and the value of C1 are calculated tomaintain nearly constant the gain of the optical link over aband wider than the band of the RF signal. Besides, thedimensions of L1 and the value of C1 shall be modified tomatching with the various allocation of the RF channel insidethe overall frequency range of the VCSEL. Introducing thematching network LASINT the measured value of the inputdynamic rangeD with500 m fibre is -39 dBm, corresponding toanSFDR of90 dB.Hz^2/3. With1.6 km fibre the values becomesD =-54 dBm, corresponding to anSFDR of80 dB.Hz^2/3.
• Fig.11 shows the network PHOTINT of fig.4 to be insertedin the optical link at pointb of fig.8. This networkincludes a decoupling capacitor C2 of 2.2 pF seriallyconnected to a transmission line L2, in turn connected to thegate of a low-noise FET T1, for example Mitsubishi MGF1908Bwith flat gain over than 15 GHz. The series of C2 and L2constitutes an impedance matching network of the FET T1 inputimpedance to the PIN diode output impedance. A resistance R3connects the drain of T1 to the polarisation voltage +VD, while the source is grounded. The gate of T1 is negativelypolarised by -V. Externally to PHOTINT the anode of PIN diodeis pulled-up to the inverse polarisation voltage -VP by meansof a resistance R2, serially connected to the PIN diode atthe input of the network connected with the capacitor C2.Resistance R2 is dimensioned to set the inverse current ofthe PIN diode. An external capacitor C3 couples the signal atthe output of T1 to the input of the amplifier A2 (fig.8).Transmission line L2 is a microstrip on a FR4 substrate0.3 mm thick (ε_r = 4.7) dimensioned to match, together withcapacitor C2, the input impedance of FET T1 to the outputimpedance of the PIN diode for avoiding signal reflectionsand minimising the Noise Figure of FET T1 in the meanwhile.Due to this constraint the gain G is not more flat as infig. 9a but weakly selective around the midband frequency(1.685 GHz). The microstrip L2 is 0.2 mm wide and 16 mm long,corresponding to 15 nH inductance at 1.685 GHz. Dimensions ofL2 microstrip and the value of C2 match the variousallocation of the RF channel inside the overall frequencyrange of laser VCSEL. The Low-noise FET T1 operates as atransimpedance amplifier in common source configuration whichideally supplies an output voltage V_o in proportion to thesignal current I_s through the PIN diode, independently of thesource resistance R_s and the external load resistance R_L. Thesource resistance R_s is the resistance of the PIN diode in aNorton's equivalent of the input circuit of FET T1, while theexternal load resistance R_L is the input resistance of the RFamplifier A2 (fig.8) in a Thevenin's equivalent of the outputcircuit of FET T1. Considered the importance the Noise Figureassumes in the design of the network PHOTINT, some usefulnotions on this argument are given in the following. A sounddefinition of the Noise Figuren_F can be done in terms ofSignal-to-Noise ratio (S/N) degradation produced by a 2-port network (the network PHOTINT). To say, the ratio of theavailable S/N power-ratio at the input of the 2-port networkto the available S/N power-ratio at the output, when thetemperature of the input noise source is 290° K. This isexpressed as:nF =psi/pnipso/pno =psi/kTo dfpso/pnowhere:
• p_si = the available signal power at the input;
• p_so = the available signal power at the output;
• p_ni = the available noise power at the input in a smallbanddf (1 Hz). Since the temperature isstandard:p_ni =kT_o df, wherek is the Boltzmann's constant(1.3805 × 10-23 joule/K);
• p_no = the available noise power at the output in a smallbanddf (1 Hz).
Using the definitiong_a(f) =p_so /p_si, whereg_a(f) is theavailable gain of the 2-port network with the driving-sourcenoise voltage and the internal noise uncorrelated, theexpression (4) is equivalent to:nF =pnog(f)kTo dfThe average noise figure is:nF =pntgo kTo Bwwherep_nt is the total available noise power at the output,g_o is the available midband gain, andB_w is the noisebandwidth including the RF channel band. The smaller theNoise Figure, the better is the sensitivity of the network todiscriminate very low input signals from noise. Expression(6) shows that the noise figure could be effectivelyminimised by introducing some restrictions on the bandwidthB_w and the gaing_o. The pure reactive impedance matchingnetwork L2, C2 at the input of PHOTINT doesn't introducenoise terms into the (5) and (6), thus is suitable forminimising the noise figuren_F . Joinedn_F minimisation andimpedance matching can be performed either experimentally orby using known mathematical algorithms. Once minimisation iscomplete both then_F and the impedance functionZ₂(f, L2, C2) is known. In such a caseZ₂(f, L2, C2) can besynthesised with known methods giving the above indicatedvalues of L2 and C2.
• The successive Figures 12a, 12b and 13a, 13b show asmany plots of the gain vs. frequency of the DUT of fig.8upgraded with the two networks LASINT and PHOTINT for twofibre lengths. It can be appreciated that gain versusfrequency is similar to a slowly decreasing bandpass functionmuch wider than the 8 MHz channel spacing of CATV.
• Fig.12a shows the gain vs. frequency measured on 500 moptical link, whilefig.12b shows the enlarged central partof the preceding plot in a band of about 100 MHz. It can beappreciate that the gain is nearly flat inside the 5 MHzchannel band. The measured value of the input dynamic range Dwith a500 m fibre is-33 dBm, corresponding to anSFDR of94dB.Hz^2/3.
• Fig.13a shows the gain vs. frequency measured on1.6 kmoptical link, whilefig.13b shows the enlarged central partof the preceding plot in a band of about 100 MHz. It can beappreciate that the gain is nearly flat inside the 5 MHzchannel band. The measured value of the input dynamic range Dwith1.6 km fibre is-49.5 dBm, corresponding to anSFDR of83 dB.Hz^2/3.
• Fig.14 schematises the network PHC of fig.4 to beinserted in the optical link at point c offig.8. Thisnetwork includes an envelope detector ENVDET and a variable phase shifter VARAC. The envelope detector ENVDET includes ahalf-wave rectifier reproducing the envelope of the modulatedRF signal present at the output of the power amplifier A2(fig.8). The rectified voltage is used as a control signal ofthe variable phase shifter VARAC. In a non-limitingembodiment a varactor diode (inversely polarised) is used asa voltage-controlled phase shifter. For a certain extent ofthe control signal level the inverse capacity of the varactoris a linear function of the control signal, so as the phaseoffset introduced into the RF output signal. The sign ofphase shift introduced by the network PHC being opposite tothe AM/PM phase distortion in correspondence of the revealedenvelope, this takes care of all the AM/PM distortions alongthe optical link.
• Fig.15a shows (solid line) the frequency spectrum of aWCDMA signal (translated at 1.680 GHz for convenience) at theend of 500 m optical link of fig.8 completed with LASINT,PHOTINT, and PHC networks. The measured value of the inputdynamic rangeD with a500 m fibre is-27 dBm, correspondingto anSFDR of98 dB.Hz^2/3. On the same Figure the spectrumwithout compensation of the phase distortion (dashed line) issuperimposed to the preceding spectrum for comparison. It canbe appreciate the lowering of the out of band interferencesachieved by the phase distortion compensation network PHC.The interference power inside adjacent channels is reducedconsequently.Fig.15b is similar to fig.15a but is referredto the comparison of the canonical ACPR (Adjacent ChannelPower Ratio) measure performed on the RRC filter bandwidth(5 MHz).
• Figures 16a and16b are similar to the Figures 15a and15b but with1.6 km fibre. The measured value of the inputdynamic rangeD with1.6 km fibre is-45 dBm, correspondingto anSFDR of86 dB.Hz^2/3.
• Additional increasing of the SFDR is possible by theintroduction of one or two gain linearisation networks inappropriate points of the optical links. A suitablelineariser is the one disclosed in the European patentEP 0451909-B1 (corresponding to US 5,138,275) titled:"Predistortion lineariser for microwave power amplifiers",inventors Carlo Buoli et al., in the name of the sameAssignee (SIEMENS) of the present invention, and incorporatedby reference. An independent claim (US) recites: A finalmicrowave power amplifier having compensation for gaincompression and phase distortion thereof, comprising:
• a transistor amplifier;
• subpolarisation biasing means connected to an input ofsaid transistor amplifier for providing gain expansionwith increase in power of the signal at the input of thetransistor amplifier by setting a subpolarisationoperating point thereof; and
• phase shifter means controlled by a continuous componentof voltage present at an output of said amplifiertransistor for introducing a phase distortion foroffsetting the phase distortion of said final poweramplifier.
• This known solution is highly profitable because makesuse of a unique common source GaAsFET transistor polarisednear pinch-off to act both as a final RF power amplifier withgain expansion for recovering amplitude distortion andcommand signal generator for a varactor used for offsettingthe phase distortion of the power amplifier. A solution withthe lineariser (both FET and varactor) placed upstream aseparate power amplifier is also claimed in the mentionedpatent. The peculiarity of this "predistortion lineariser" isthat the two type of distortions AM/AM and AM/PM arecompensated together, so that a disjoint variable phase shifter equivalent to the PHC network (fig.14) is notachievable, and the sole compensation of the AM/PM distortionimpossible. The contrary is not completely true and the soleAM/AM distortion can be compensated by omitting the varactorcircuitry. There are not indications in the disclosure of thecited patent about a possible use of the "predistortionlineariser" to build-up an analog optical link. This knownlineariser appositely thought for a microwave power amplifieris now forced to further compensate for the AM/PM distortiondue to the intermodal dispersion inside the optical fibre.With reference to fig.8 the following are possible use of theknown "predistortion lineariser":
• a first simplest and low cost embodiment, particularlyindicated for low amplifications and short fibre lengths,is that to use the predistortion lineariser in place ofFET T1 (fig.11) and do not use the amplifier A2.
• A second embodiment, coinciding with the second variant,of the invention is that to use the predistortionlineariser in place of the amplifier A2 and thecompensating network PHC.
• A third embodiment, coinciding with the third variant, isthat to replace the amplifier A1 with the said lineariserdeprived of the varactor.

Claims (10)

1. Analog optical link including the following elements:

   a first RF amplifier (A1) of an analog bandpass RF signal;

   a first impedance matching network (LASINT) to match theinput impedance of a downstream VCSEL laser to the outputimpedance of the first RF amplifier (A1);

   said VCSEL laser coupled to one end of a multimode opticalfibre of given length for transducing the electricalbandpass RF signal into an optical signal injected intothe multimode fibre;

   a PIN diode coupled to the other end of the multimodeoptical fibre for transducing into electrical the opticalsignal coming from the fibre;

   a second RF amplifier (T1) of transimpedance type;

   characterised in that said analog optical link includes:

   a second impedance matching network (C2, L2) adapted tomatch the input impedance of the transimpedance amplifier(T1) to the output impedance of the PIN diode, minimisingin the meanwhile the noise figure of said transimpedanceamplifier (T1) upon a frequency range including thebandpass of the RF signal;

   an amplitude envelope detector (ENVDET) placed downstreamthe transimpedance amplifier (T1) for generating anenvelope control signal;

   a variable phase shifter (VARAC) controlled by saidenvelope control signal to introduce a phase offset on theRF signal at the output of the transimpedance amplifier(T1) compensating for the AM/PM distortion introduced bythe optical fibre.
2. Analog optical link of the claim1,characterised inthat includes a third RF amplifier (A2) placed between saidtransimpedance amplifier (T1) and said variable phase shifter(VARAC).
3. Analog optical link of the claim 2,characterised inthat the third RF amplifier (A2) is a final power amplifiersupplying a transmitting antenna and the link includes a gainlinearisation network of said third RF power amplifier tocompensate for the AM/AM distortion introduced on the powerpeaks of the RF signal.
4. Analog optical link of the claim 3,characterised inthat the third RF amplifier (A2) includes subpolarisationmeans of a power transistor constituting a gain expansionstadium operating as predistortion lineariser furthergenerating said envelope control signal for said variablephase shifter (VARAC).
5. Analog optical link of one of the preceding claims,characterised in that includes a gain linearisation networkdesigned to compensate for the AM/AM distortion of the firstRF power amplifier (A1) on the power peaks of the RF signalin combination with the highest amplifications needed todrive the VCSEL adequately for the longest optical links.
6. Analog optical link of one of the preceding claims,characterised in that the multimode optical fibre is up to150 m long to match up to typical lengths of theintra-building connections involved in CATV service, or toconnect indoor apparatuses with a tower antenna, inparticular an antenna array suitable for beamforming.
7. Analog optical link of one of the preceding claimsexcept the preceding,characterised in that the multimodeoptical fibre is up to 500 m long to match up to a small-sizefibre infrastructure suitable for connecting a centralisedbase station of a pico/microcellular system to a pool ofremote antennas serving as many cells.
8. Analog optical link of one of the preceding claimsexcept the preceding one,characterised in that the multimodeoptical fibre is up to 1.6 km long to match up to a fibre infrastructure suitable for connecting a centralised basestation of a cellular system either of the second or thirdgeneration to a pool of remote antennas.
9. Analog optical link of one of the preceding claims,characterised in that said first impedance matching network(LASINT) includes a discrete capacitor (C1) seriallyconnected to a microstrip (L1), being the value of saidcapacitor (C1) and the dimensions of said microstrip (L1)calculated to maintain nearly constant the gain of theoptical link over a band wider than the band of the RFsignal.
10. Analog optical link of one of the preceding claims,characterised in that said second impedance matching network(C2, L2) includes a discrete capacitor (C2) seriallyconnected to a microstrip (L2).

Images